Compact discs (CD's) on which digital audio signals such as those of music are recorded have gained widespread use. Also introduced are compact disc read-only memories (CD-ROM'S) accommodating not only music data but also image- and computer-related data. Today, there are signal formats extended from the CD-ROM such as CD-ROM-XA (CD-ROM extended architecture) and CD-I (CD-interactive). The CD-ROM-XA format is an extended CD-ROM format that minutely defines how audio and video data are to be recorded. The CD-I format is a format that emphasizes the interactive operation between the medium and users. The audio information complying with the CD-ROM-XA format and the CD-I format is recorded onto discs by adaptive differential pulse code modulation (ADPCM).
FIG. 1 (A) shows a typical data structure of a CD-ROM-XA disc. Illustratively, one track may record up to 255 files of a given size. When audio data are to be recorded, each file may be divided into 16 channels each for use independent of one another, as depicted in FIG. 1 (B). Where image data, text data or computer programs are to be recorded, each file may be divided into 32 channels each for use independent of one another, as indicated in FIG. 1 (C).
FIGS. 2 and 3 show respectively six signal formats and six recording formats in which to record audio data to each file of the CD-ROM-XA disc.
Format a is for stereo recording. With this format, the original stereo audio signal is subjected to A/D conversion at a sampling frequency of 37.8 kHz and with a quantization bit count of 16. The resulting audio data in digital format are compressed by ADPCM into eight bits per sample. As format a in FIG. 3 indicates, the compressed data are recorded on an interleaving basis to every other block in a plurality of blocks (sectors) making up each file.
With format b, the original monaural audio signal is subjected to A/D conversion at a sampling frequency of 37.8 kHz and with a quantization bit count of 16. The resulting audio data in digital format are compressed by ADPCM into eight bits per sample. As format b in FIG. 3 shows, the compressed data are recorded on an interleaving basis to every fourth block in a plurality of blocks constituting each file.
The same workings apply to formats c through f. With any of these formats, the original audio signal is encoded into ADPCM audio data. The resulting audio data are recorded on an interleaving basis in each of the formats c through f.
For normal use, the period per block is 1/75 sec., and the amount of data that may be recorded per block is 2,324 bytes.
Because audio data are recorded on an interleaving basis as depicted in FIG. 3, other data may be recorded to unused blocks. For example, where format a is used to record audio data in interleaving fashion, four-channel stereo recording is available. That is, four languages may be included in a single program. If image data are additionally recorded, the audio signal may be played back accompanied by image reproduction.
Typical image data that may be recorded on CD-ROM-XA discs will now be described with reference to FIGS. 4 through 10.
FIG. 4 is a flowchart of steps in which to record image data in units of frames. FIG. 5 (A) shows a typical image data screen construction for a single frame. The image data for one frame are made of 256 pixels (in traverse direction) by 192 pixels (in longitudinal direction). The red, green and blue for each pixel are represented by five bits each. Each pixel is constituted by 16 bits, i.e., 15 bits (5 bits.times.3) plus one dummy bit.
In step 81 of FIG. 4, the original image data of one frame shown in FIG. 5 (A) are divided into blocks each measuring 8.times.8 pixels (each block is called a character), as illustrated in FIG. 5 (B). That is, the image of one frame is split into 768 characters (32 .times.24) CHR(0) through CHR(767). The data on one character CHR(i)(i=0-767) are made of 128 bytes (8 pixels.times.8 pixels.times.16 bits).
In step 82, primary vector quantization is performed on the character data CHR(i). The vector quantization executed here involves quantizing the character data so that the number of pixel colors within the character will be limited to a maximum of four. Specifically, in this example, image data are created as follows: A three-dimensional color space is first considered in which the color components of red, green and blue are taken on three coordinate axes that intersect one another orthogonally. In this color space, the distances between the pixels are obtained. Then the pixels whose distances to one another are sufficiently short are grouped together so that the pixel colors within one character are limited to four or fewer representative colors.
The above vector quantization is allowed to continue immediately before a maximum quantization error Emax of each character (i.e., distance between representative color and pixel) is exceeded for each character. The process makes uniform the S/N ratios of all characters (C0 through C767) in each frame. The character data CHR(i) are quantized in this manner, thereby limiting the number of colors in each character to a maximum of four.
In step 83, the quantized character data CHR(i) are categorized into eight groups each comprising characters of like color tones (each group is called a palette). The character data CHR(i) for each character are grouped into one of eight palettes P0 through P7.
In the grouping above, the characters are simply grouped and their sequence is not altered. There is no need for a given palette Pj (j=0 to 7) to be made of contiguous character areas. That is, characters located in a scattered manner may constitute one palette. For example, as shown in FIG. 5 (C), areas A through E of like color tones may each form part of the palette Pj.
In step 84, secondary vector quantization is carried out on the character data CHR(i) grouped into each palette Pj. Even though each character Ci has four or fewer representative colors, the palette Pj comprising characters Ci may have more than 16 colors. If one palette Pj has more than 16 colors, the palette Pj is subjected to secondary vector quantization in the same manner as primary vector quantization so that the number of colors within the palette Pj will be limited to a maximum of 16. The character data CHR(i) belonging to each palette Pj are quantized into the color data (of 15 bits) about one of the 16 representative colors specific to that palette.
In step 85, color number conversion tables COL(j) are created for the palettes Pj. As shown in FIG. 6, the table COL(j) for each pallet Pj is a conversion table that contains the color data (in 15 bits) about the 16 representative colors specific to that palette as well as four-bit color numbers (0-15) for designating the color data. At this point, the character data CHR(i) are equal to one set of character data in the table COL(j) of the palette to which the character data CHR(i) belong.
In step 86, the character data CHR(i) about the respective colors following secondary vector quantization are converted to color numbers in the color number conversion table COL(j) of the palette Pj to which the character CHR(i) belong. The conversion is carried out by reference to the table COL(j). The pixels in each character are represented by two parameters: a four-bit color number designating each color, and data indicating the color number conversion table COL(j) to which the color number belongs.
In step 87, a screen table SCR is created. As illustrated in FIG. 8, the screen table SCR has a total of 768 addresses corresponding to 24.times.32 characters per frame of the original image data. Each address is two bytes long, as shown in FIG. 9. Of the two bytes, the low-order 10 bits constitute a number Ci indicating a character. The high-order three bits make up a palette number Pj designating the palette Pj to which the character data CHR(i) associated with the character number Ci belong.
In step 88, the character number Ci of the screen table SCR is shifted by 16 in the high-order direction, with character numbers 0 through 15 assigned to the color number of a monochromatic character Ci. For the monochromatic character Ci, the palette number Pj remains the palette number specific to that character but the character number Ci designates the color number. At this point, the address of the monochromatic character Ci in the table SCR indicates the character number Ci.
In the manner described, image data are converted per frame to a color number conversion table COL(j), to a screen table SCR, and to a four-bit color number for each pixel. In the description that follows, the color number conversion table COL(j) will be called the color number conversion table COL, and the four-bit color number for each pixel will be referred to as the color number data PAT. The data PAT, table SCR and table COL may illustratively be combined to form recording data RECD in the format shown in FIG. 10. The recording data are recorded on the CD-ROM-XA disc after undergoing a predetermined encoding process.
The amount of image data RECD recorded per frame is calculated as follows: Eight palettes Pj exist per frame. One palette has 16 colors, each color being represented by 16 bits (of which 1 bit is dummy). Thus the color number conversion table COL is made of a total of 256 bytes (=8 palettes.times.16 colors.times.16 bits). Since there are 768 characters and each character corresponds to two bytes, the screen table SCR amounts to 1,536 bytes (=768 characters.times.2 bytes). With the color number data DAT comprising four bits per pixel, the amount of data DAT per character is 32 bytes (4 bits .times.64 pixels). Because the data about a monochromatic character Ci are already transmitted as part of the screen table SCR, it is not necessary to transmit the data anew. Thus if a quarter of the 768 characters per frame are monochromatic characters and the remaining three fourths are multi-colored characters, the total data amount is obtained by adding 256 bytes of the color number conversion table COL, 1,536 bytes of the screen table SCR, and 18,432 bytes of the color number data DAT (=32 bytes.times.768 characters.times.3/4). The sum is 20,224 bytes.
Where image data RECD are to be recorded on the CD-ROM-XA disc, all 32 channels may be used to record the data, as shown in FIG. 1 (C). At this time, the transmission rate of the CD-ROM-XA disc for the image data RECD is 174,300 bytes (=2,324 bytes.times.75 blocks) per second. In this example, the frame count per second is about 8.6 (=174,300 bytes divided by 20,224 bytes). Animated images are recorded and reproduced with this frame count.
As illustrated in FIG. 1, each file on the CD-ROM-XA disc may record audio data and image data mixedly in units of blocks. Where a computer game is executed from recordings on a CD-ROM-XA disc, background music (BGM) and sound effects may be reproduced concurrently to enhance the entertainment value of the game. Illustratively, format e in FIG. 2 or 3 is utilized so that one out of eight blocks will address audio data and the remaining seven will address image data RECD. In this manner, both music and animated images are reproduced at the same time.
Because seven out of eight blocks are used to record and reproduce image data RECD in the above example, the frame count per second for the image data RECD is about 7.5 (8.6 frames.times.7/8). That frame count is insufficient for the adequate display of animated images. This drawback is circumvented illustratively by raising the revolutions of the disc. Rotating at higher speeds, the disc increases the number of image frames per second for animation. Illustratively, doubling the standard revolutions of the disc doubles the rate of data reproduction therefrom. That is, with the standard disc rotating speed doubled, the frame count is about 15 (=7.5 frame.times.2) per second, which is sufficient for animated image reproduction. Descriptions of how data are recorded and reproduced to and from a disc at a rotating speed higher than usual are found illustratively in a pending international application, "Data Recording and Reproducing Method" (by Yoichiro Sako, Serial No. PCT/JP91/00054, filed Jan. 18, 1991).
As described, concurrent recording of audio data and animated image data on a disc should preferably be done by raising the disc revolutions (e.g., twice as fast as standard rotating speed). This scheme involves one disadvantage. That is, to double the disc revolutions requires lowering in half the frequency of the original audio signal as the signal is subjected to ADPCM for audio data preparation. Optimum encoding of the audio signal is disabled at this point, with the result that the sound quality of the reproduced audio signal deteriorates.